Article’s

The prompt and precise identification of heart abnormalities from biological information is essential in advanced healthcare systems. This paper introduces an AI-based hybrid framework for real-time detection of cardiac anomalies by combining electrocardiogram (ECG) and photoplethysmogram (PPG) signal processing with sophisticated machine learning methodologies. The suggested system employs extensive preprocessing and augmentation techniques, such as Gaussian noise injection, amplitude scaling, and temporal shifting, to increase signal diversity and enhance model generalization robustness. Morphological and temporal cardiac characteristics—including P-wave duration, PR interval, QRS complex width, ST-segment level, T-wave duration, and Pulse Transit Time (PTT)—are obtained utilizing the WFDB, NeuroKit2, and BioSPPy frameworks. Annotation-assisted feature labeling and automated P-wave delineation are integrated to guarantee dependable beat-level characterisation. An ensemble CatBoost model is utilized for classification, exhibiting enhanced efficacy compared to traditional Random Forest classifiers in managing non-linear, multi-dimensional biological data. The model’s efficacy is assessed by cross-validation and confusion matrix analysis, resulting in a mean accuracy enhancement over 15% relative to baseline approaches. The findings underscore the efficacy of gradient boosting topologies for comprehensive cardiac health evaluation. This framework establishes a basis for real-time, AI-enhanced heart monitoring and can be further incorporated into smart wearable and telemedicine systems to enable early detection and predictive diagnosis in cardiovascular healthcare